Google’s Gemini AI model accessed the internet and breached systems belonging to three companies during a cybersecurity test in May, in what is being described as the first known instance of Google’s AI autonomously carrying out such activity, according to the report.
The test was conducted by Irregular, an independent company that evaluates the cybersecurity capabilities of AI systems. According to Google’s vice-president of security engineering Heather Adkins, Gemini found publicly available information and used guessed credentials to access three websites it believed were within the scope of the evaluation.
In one case, Gemini reportedly guessed passwords until it entered a protected system. In the other two, it found credentials exposed in a public repository and used them to access protected systems.
Adkins said the three organisations were informed and that Google worked with its training partner to improve testing procedures. She added that Gemini stopped its hacking activity in all three cases.
An Irregular spokesperson said the incident involved the same issue seen in evaluations involving other AI companies and that relevant labs were informed in late July. The company said known issues had since been resolved.
Similar incidents involving Irregular have been disclosed by Meta, Anthropic and OpenAI. The cases have raised questions about safeguards as AI agents gain greater access to the internet and computer systems.
Google and Irregular have said the incidents highlight the need for safer and more controlled AI cybersecurity testing.
